Currently funded project provided data on 111 schizophrenic and schizoaffective patients assigned at random to haloperidol (HAL) treatment at fixed plasma levels ranging from 2 to 34 ng/ml. No relationship between plasma levels and clinical efficacy has emerged. This result suggests that the current clinical practice may be wrong; patients are apparently receiving higher doses of HAL than they need. This is an important problem, particularly because of the HAL side effects. The primary goal of this continuing project is to determine the clinical efficacy of very low plasma levels of HAL in schizophrenia. The subjects will be acutely exacerbating schizophrenics newly admitted to inpatient services. a total of 180 subjects will be treated with one of the two plasma level HAL ranges: "low" (0.5-3.5 ng/ml), or "high" (8.5-11.5 ng/ml). Lorazepam and benztropine will be available as concomitant medications if needed. The subjects will be maintained in their randomly assigned HAL range for 3 weeks. Those who fail to improve at the end of this period will be randomly assigned to either the same level or to the other HAL level range for another 3 weeks. The patient evaluations will include the BPRS, SAPS, SANS, and Simpson-Angus Scale. Blood assays for HAL will be used for dose adjustments. All evaluations and assays will be double-blind. The principal three hypothesis to be tested: 1. High and low levels of HAL will yield similar clinical improvement at the endpoint of the first three week period; 2. The subjects who fail to respond to the low level of HAL during the first 3 weeks will not appreciably benefit from the switch to the high level. 3. The two levels of HAL will have a similar speed of onset of clinical effect.